Research On Delay Compensation And Non-steady Flatness Intelligent Control For Cold Rolled Strip

The flatness control accuracy of ordinary strip is already very high in the steady rolling stage,but for high value-added products of cold rolled thin strip,the problem of flatness is still prominent in the steady rolling stage.In addition,the proportion of strip quality shear losses,quality objections,and shutdowns caused by excessive flatness quality fluctuation in the non-steady rolling process have exceeded 1/3 in the non-equipment failure shutdown.These issues have also become an adverse factor restricting the enterprises to obtain greater profits.In this thesis,the 1450 mm UCMW five-stand tandem cold rolling mill in a factory is taken as the research object,and the system delay compensation in the steady-state rolling stage and the flatness quality problem in the non-steady rolling process are deeply studied.The main research contents are as follows:(1)The flatness data is collected from steady and non-steady rolling processes,which clearly indicates the flatness quality issue in the cold rolling process.Smith predictive controller is introduced to compensate the flatness control system,and the system models of the work roll bending,the intermediate roll bending,the roll tilting,and the intermediate roll shifting are established respectively.Due to the different installation positions of the work roll shifting compared to other flatness actuators,based on the metal flow principle,the calculation method was provided for its control system delay time,and the system model of the work roll shifting is also established.Through simulation experiments,the dynamic quality of the work roll bending control system under conventional PID,Smith predictive PID,and Smith predictive fuzzy PID control methods during steady rolling was compared.Based on the calculated delay time of the single taper work roll shifting control system,a predictive model was established to compensate for this large delay control system.The effectiveness of the model predictive control are confirmed by the comparison of theoretical analysis,simulation experiments and application effect.(2)Two different deep neural networks are established to update the Actor and Critic network.The output of RL Agent actions is optimized to make the performance of DDPG intelligent algorithm more stable.The model of the non-steady flatness DDPG intelligent control system is also established.Based on the non-steady flatness data of acceleration and deceleration in actual production,the simulated non-steady interference model was used in the developed DDPG intelligent algorithm.By training the AL Agent and simulating the working roll bending system model with different flatness actuators efficiency factors,it was verified that the DDPG intelligent algorithm can effectively weaken the interference during the non-steady rolling stage,the insensitivity to the parameters of the controlled object During the nonsteady rolling stage,the ideal control effect have been achieved on the flatness quality.a flatness control strategy for the steady and non-steady rolling stages of the 1450 mm UCMW five stand tandem cold rolling mill is formulated.